Model Card for Model ID
This project detects whether a given text is human-written, AI-generated, or paraphrased using state-of-the-art NLP models.
Model Description
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
Dataset
Training Data The model was trained on 46,181 text samples. Data sources include: Human-Written: Web-scraped research papers & verified content. AI-Generated: Text generated by ChatGPT-4. Paraphrased: Text rewritten using Grammarly & QuillBot.
Model Sources [optional]
Uses
⚠ Disclaimer This model is designed to assist in identifying whether a text is human-written, AI-generated, or paraphrased, but it should not be solely relied upon for critical decisions. It may misinterpret text, produce false positives/negatives, or struggle with highly modified AI-generated content.
For best results, use this model as a supporting tool, and always verify its outputs manually.
Direct Use
This model is designed to classify text as Human-Written, AI-Generated, or Paraphrased. It can be used in education, content moderation, and plagiarism detection.
Out-of-Scope Use
❌ Not suitable for legal or forensic content verification.
Bias, Risks, and Limitations
⚠ Potential Bias: The model is trained on limited datasets, which may not generalize well to non-English texts or niche writing styles. ⚠ False Positives: AI-generated and paraphrased content may sometimes be misclassified. ⚠ Adversarial Attacks: The model may not detect AI-generated text if rewritten with subtle human edits.
Recommendations
Users should not rely solely on this model for high-stakes decision-making. Always verify outputs manually before taking action.
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
How to Get Started with the Model
Use the code below to get started with the model.
First install dependencies: pip install transformers torch
Then, load the model using:
from transformers import AutoModelForSequenceClassification, AutoTokenizer
model_name = "your-username/ai-content-classifier"
model = AutoModelForSequenceClassification.from_pretrained(model_name) tokenizer = AutoTokenizer.from_pretrained(model_name) def classify_text(text): inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=512) with torch.no_grad(): outputs = model(**inputs) logits = outputs.logits predicted_class = torch.argmax(logits, dim=1).item() labels = {0: "Human-Written", 1: "AI-Generated", 2: "Paraphrased"} return labels[predicted_class]
text = "This is a sample AI-generated text." print(classify_text(text))
Training Details
Preprocessing [optional]
Preprocessing [optional]
Tokenization: AutoTokenizer from Hugging Face. Data Cleaning: Removed duplicates, special characters, excessive whitespace.
Training Hyperparameters
- Training regime: [More Information Needed]
Base Model: ELECTRA Learning Rate: 0.00002 - 0.00003 Batch Size: 8 - 16 Epochs: 3 Optimizer: AdamW Max Token Length: 512